Gilder argued that just as the microprocessor had introduced previously unimaginable processing power, so the fiber-optic construction boom would usher in a world of instantaneous communication and infinite bandwidth: the telecosm. He predicted that it would make “the CPU … peripheral, the network central,” and that it would enable anyone to launch a product, company, or political movement. But every boom must go bust, and the crash of the telecommunications industry, when it came, proved worse than the bursting of the dot-com bubble. More than $500 billion was lost in just a few years. Between 2001 and 2004, 216 telecommunications companies went bankrupt–most notably Worldcom ($104 billion in assets), whose CEO, Bernie Ebbers, received a 25-year jail sentence for what remains the largest accounting fraud in U.S. history. Meanwhile, hitherto stable industry giants like AT&T staggered. Unfortunately for Gilder, he had loved his tech companies not wisely but too well, investing his own money as he had advised others to do.
“I’m a fan of George Gilder, the bubble bursting notwithstanding,” Ethernet coinventor Bob Metcalfe (a member of Technology Review’s board of directors) told me after his San Diego keynote speech, “Toward Terabit Ethernet.” Metcalfe had told his audience not only that optical networks would soon deliver 40- and 100-gigabit-per-second Ethernet–standards bodies are now hammering out the technical specifications–but also that 1,000-gigabyte-per-second Ethernet, which Metcalfe dubbed “terabit Ethernet,” would emerge around 2015. Why, I asked, did Metcalfe believe this? “Last night, Gilder spoke to 300 of us at an executive forum about his ‘Exaflood’ paper, in which he predicts a zettabyte of U.S. Internet traffic by the year 2015,” Metcalfe said. “Since I admire Gilder, I extrapolated from his prediction.”
An exabyte is 1018 bytes of data; a zettabyte is 1021 bytes. Metcalfe pointed to video, new mobile, and embedded systems as the factors driving this rising data flood: “Video is becoming the Internet’s dominant traffic, and that’s before high definition comes fully online. Mobile Internet just passed a billion new cell phones per year. Then totally new sources of traffic exist, like the 10 billion embedded microcontrollers now shipped annually.” Did Metcalfe believe that the existing infrastructure–built in the boom years, when great excesses of fiber-optic cable were laid down–could support terabit Ethernet? “That dark fiber laid down then is being lit up, and some routes are now full,” he said. “That’s the principal pressure to go to 40 and 100 gigabits per second. It seems we can reach those speeds with basically the same fibers, lasers, photodetectors, and 1,500-nanometer wavelengths we have, mostly by means of modulation improvement. But it’s doubtful we’ll wring another factor of 10 beyond that.” Thus, the backbone networks would need to be overhauled and new technologies implemented.
The speaker after Metcalfe, Herwig Kogelnik, described both the field’s progress and the technologies that would support not just 10- and 40-gigabyte-per-second but also terabit speeds. Kogelnik–who in more than four decades at Bell Labs has headed several research divisions investigating lasers, holography, and optical guided-wave devices, collecting too many academic and industry honors to list in less than a page–explained that current research had, for example, advanced WDM (wavelength division multiplexing) technology to a point where economical transmission of 10 channels, each carrying 100-gigabyte-per-second traffic, was now feasible. Likewise, on the trade-show floor, it was apparent that the component technologies of the telecosm Gilder envisioned a decade ago–a global network with infinite bandwidth and instantaneous transmission–were becoming available in 2008. Companies exhibited products that made use of silicon photonics: Lightwire, for instance, offered a lightweight transceiver designed to greatly improve upon the SFP+ modules currently used to connect servers and network equipment. Since photons move much faster and scatter much less heat than electrons, it promises to reduce power dissipation by more than half.